Cluster identification and separation in the growing self-organizing map: application in protein sequence classification

Growing self-organizing map (GSOM) has been introduced as an improvement to the self-organizing map (SOM) algorithm in clustering and knowledge discovery. Unlike the traditional SOM, GSOM has a dynamic structure which allows nodes to grow reflecting the knowledge discovered from the input data as learning progresses. The spread factor parameter (SF) in GSOM can be utilized to control the spread of the map, thus giving an analyst a flexibility to examine the clusters at different granularities. Although GSOM has been applied in various areas and has been proven effective in knowledge discovery tasks, no comprehensive study has been done on the effect of the spread factor parameter value to the cluster formation and separation. Therefore, the aim of this paper is to investigate the effect of the spread factor value towards cluster separation in the GSOM. We used simple k-means algorithm as a method to identify clusters in the GSOM. By using Davies–Bouldin index, clusters formed by different values of spread factor are obtained and the resulting clusters are analyzed. In this work, we show that clusters can be more separated when the spread factor value is increased. Hierarchical clusters can then be constructed by mapping the GSOM clusters at different spread factor values.

Mining for Class-Specific Motifs in Protein Sequence Classification

Background: In protein sequence classification, identification of the sequence motifs or n-grams that can precisely discriminate between classes is a more interesting scientific question than the classification itself. A number of classification methods aim at accurate classification but fail to explain which sequence features indeed contribute to the accuracy. We hypothesize that sequences in lower denominations (n-grams) can be used to explore the sequence landscape and to identify class-specific motifs that discriminate between classes during classification. Discriminative n-grams are short peptide sequences that are highly frequent in one class but are either minimally present or absent in other classes. In this study, we present a new substitution-based scoring function for identifying discriminative n-grams that are highly specific to a class. Results: We present a scoring function based on discriminative n-grams that can effectively discriminate between classes. The scoring function, initially, harvests the entire set of 4- to 8-grams from the protein sequences of different classes in the dataset. Similar n-grams of the same size are combined to form new n-grams, where the similarity is defined by positive amino acid substitution scores in the BLOSUM62 matrix. Substitution has resulted in a large increase in the number of discriminatory n-grams harvested. Due to the unbalanced nature of the dataset, the frequencies of the n-grams are normalized using a dampening factor, which gives more weightage to the n-grams that appear in fewer classes and vice-versa. After the n-grams are normalized, the scoring function identifies discriminative 4- to 8-grams for each class that are frequent enough to be above a selection threshold. By mapping these discriminative n-grams back to the protein sequences, we obtained contiguous n-grams that represent short class-specific motifs in protein sequences. Our method fared well compared to an existing motif finding method known as Wordspy. We have validated our enriched set of class-specific motifs against the functionally important motifs obtained from the NLSdb, Prosite and ELM databases. We demonstrate that this method is very generic; thus can be widely applied to detect class-specific motifs in many protein sequence classification tasks. Conclusion: The proposed scoring function and methodology is able to identify class-specific motifs using discriminative n-grams derived from the protein sequences. The implementation of amino acid substitution scores for similarity detection, and the dampening factor to normalize the unbalanced datasets have significant effect on the performance of the scoring function. Our multipronged validation tests demonstrate that this method can detect class-specific motifs from a wide variety of protein sequence classes with a potential application to detecting proteome-specific motifs of different organisms.

Present perspectives on the automated classification of the G-protein coupled receptors (GPCRs) at the protein sequence level

The G-protein coupled receptors--or GPCRs--comprise simultaneously one of the largest and one of the most multi-functional protein families known to modern-day molecular bioscience. From a drug discovery and pharmaceutical industry perspective, the GPCRs constitute one of the most commercially and economically important groups of proteins known. The GPCRs undertake numerous vital metabolic functions and interact with a hugely diverse range of small and large ligands. Many different methodologies have been developed to efficiently and accurately classify the GPCRs. These range from motif-based techniques to machine learning as well as a variety of alignment-free techniques based on the physiochemical properties of sequences. We review here the available methodologies for the classification of GPCRs. Part of this work focuses on how we have tried to build the intrinsically hierarchical nature of sequence relations, implicit within the family, into an adaptive approach to classification. Importantly, we also allude to some of the key innate problems in developing an effective approach to classifying the GPCRs: the lack of sequence similarity between the six classes that comprise the GPCR family and the low sequence similarity to other family members evinced by many newly revealed members of the family.

Structural Alignment based Kernels for Protein Structure Classification

Structural alignments are the most widely used tools for comparing proteins with low sequence similarity. The main contribution of this paper is to derive various kernels on proteins from structural alignments, which do not use sequence information. Central to the kernels is a novel alignment algorithm which matches substructures of fixed size using spectral graph matching techniques. We derive positive semi-definite kernels which capture the notion of similarity between substructures. Using these as base more sophisticated kernels on protein structures are proposed. To empirically evaluate the kernels we used a 40% sequence non-redundant structures from 15 different SCOP superfamilies. The kernels when used with SVMs show competitive performance with CE, a state of the art structure comparison program.

Decreasing the number of false positives in sequence classification

Abstract Background A large number of probabilistic models used in sequence analysis assign non-zero probability values to most input sequences. To decide when a given probability is sufficient the most common way is bayesian binary classification, where the probability of the model characterizing the sequence family of interest is compared to that of an alternative probability model. We can use as alternative model a null model. This is the scoring technique used by sequence analysis tools such as HMMER, SAM and INFERNAL. The most prevalent null models are position-independent residue distributions that include: the uniform distribution, genomic distribution, family-specific distribution and the target sequence distribution. This paper presents a study to evaluate the impact of the choice of a null model in the final result of classifications. In particular, we are interested in minimizing the number of false predictions in a classification. This is a crucial issue to reduce costs of biological validation. Results For all the tests, the target null model presented the lowest number of false positives, when using random sequences as a test. The study was performed in DNA sequences using GC content as the measure of content bias, but the results should be valid also for protein sequences. To broaden the application of the results, the study was performed using randomly generated sequences. Previous studies were performed on aminoacid sequences, using only one probabilistic model (HMM) and on a specific benchmark, and lack more general conclusions about the performance of null models. Finally, a benchmark test with P. falciparum confirmed these results. Conclusions Of the evaluated models the best suited for classification are the uniform model and the target model. However, the use of the uniform model presents a GC bias that can cause more false positives for candidate sequences with extreme compositional bias, a characteristic not described in previous studies. In these cases the target model is more dependable for biological validation due to its higher specificity.

A clustering method for robust and reliable large scale functional and structural protein sequence annotation

Bioinformatics, in the last few decades, has played a fundamental role to give sense to the huge amount of data produced. Obtained the complete sequence of a genome, the major problem of knowing as much as possible of its coding regions, is crucial. Protein sequence annotation is challenging and, due to the size of the problem, only computational approaches can provide a feasible solution. As it has been recently pointed out by the Critical Assessment of Function Annotations (CAFA), most accurate methods are those based on the transfer-by-homology approach and the most incisive contribution is given by cross-genome comparisons. In the present thesis it is described a non-hierarchical sequence clustering method for protein automatic large-scale annotation, called “The Bologna Annotation Resource Plus” (BAR+). The method is based on an all-against-all alignment of more than 13 millions protein sequences characterized by a very stringent metric. BAR+ can safely transfer functional features (Gene Ontology and Pfam terms) inside clusters by means of a statistical validation, even in the case of multi-domain proteins. Within BAR+ clusters it is also possible to transfer the three dimensional structure (when a template is available). This is possible by the way of cluster-specific HMM profiles that can be used to calculate reliable template-to-target alignments even in the case of distantly related proteins (sequence identity < 30%). Other BAR+ based applications have been developed during my doctorate including the prediction of Magnesium binding sites in human proteins, the ABC transporters superfamily classification and the functional prediction (GO terms) of the CAFA targets. Remarkably, in the CAFA assessment, BAR+ placed among the ten most accurate methods. At present, as a web server for the functional and structural protein sequence annotation, BAR+ is freely available at http://bar.biocomp.unibo.it/bar2.0.

Protein sequence design on the basis of topology optimization techniques -Using continuous modeling of discrete amino acid types

The notion of optimization is inherent in protein design. A long linear chain of twenty types of amino acid residues are known to fold to a 3-D conformation that minimizes the combined inter-residue energy interactions. There are two distinct protein design problems, viz. predicting the folded structure from a given sequence of amino acid monomers (folding problem) and determining a sequence for a given folded structure (inverse folding problem). These two problems have much similarity to engineering structural analysis and structural optimization problems respectively. In the folding problem, a protein chain with a given sequence folds to a conformation, called a native state, which has a unique global minimum energy value when compared to all other unfolded conformations. This involves a search in the conformation space. This is somewhat akin to the principle of minimum potential energy that determines the deformed static equilibrium configuration of an elastic structure of given topology, shape, and size that is subjected to certain boundary conditions. In the inverse-folding problem, one has to design a sequence with some objectives (having a specific feature of the folded structure, docking with another protein, etc.) and constraints (sequence being fixed in some portion, a particular composition of amino acid types, etc.) while obtaining a sequence that would fold to the desired conformation satisfying the criteria of folding. This requires a search in the sequence space. This is similar to structural optimization in the design-variable space wherein a certain feature of structural response is optimized subject to some constraints while satisfying the governing static or dynamic equilibrium equations. Based on this similarity, in this work we apply the topology optimization methods to protein design, discuss modeling issues and present some initial results.

A Deterministic Optimization Approach to Protein Sequence Design Using Continuous Models

Determining the sequence of amino acid residues in a heteropolymer chain of a protein with a given conformation is a discrete combinatorial problem that is not generally amenable for gradient-based continuous optimization algorithms. In this paper we present a new approach to this problem using continuous models. In this modeling, continuous "state functions" are proposed to designate the type of each residue in the chain. Such a continuous model helps define a continuous sequence space in which a chosen criterion is optimized to find the most appropriate sequence. Searching a continuous sequence space using a deterministic optimization algorithm makes it possible to find the optimal sequences with much less computation than many other approaches. The computational efficiency of this method is further improved by combining it with a graph spectral method, which explicitly takes into account the topology of the desired conformation and also helps make the combined method more robust. The continuous modeling used here appears to have additional advantages in mimicking the folding pathways and in creating the energy landscapes that help find sequences with high stability and kinetic accessibility. To illustrate the new approach, a widely used simplifying assumption is made by considering only two types of residues: hydrophobic (H) and polar (P). Self-avoiding compact lattice models are used to validate the method with known results in the literature and data that can be practically obtained by exhaustive enumeration on a desktop computer. We also present examples of sequence design for the HP models of some real proteins, which are solved in less than five minutes on a single-processor desktop computer Some open issues and future extensions are noted.

Stochastic dynamics modeling of the protein sequence length distribution in genomes: implications for microbial evolution

In this paper, we report an analysis of the protein sequence length distribution for 13 bacteria, four archaea and one eukaryote whose genomes have been completely sequenced, The frequency distribution of protein sequence length for all the 18 organisms are remarkably similar, independent of genome size and can be described in terms of a lognormal probability distribution function. A simple stochastic model based on multiplicative processes has been proposed to explain the sequence length distribution. The stochastic model supports the random-origin hypothesis of protein sequences in genomes. Distributions of large proteins deviate from the overall lognormal behavior. Their cumulative distribution follows a power-law analogous to Pareto's law used to describe the income distribution of the wealthy. The protein sequence length distribution in genomes of organisms has important implications for microbial evolution and applications. (C) 1999 Elsevier Science B.V. All rights reserved.